you're a dev forced to shitslophow do you get out of this situation?"wait two more weeks until AI gets better and improved the code" doesn't count.
>>110280335No one understands 100% of how the human body works, but it works anyway.Codebases in the future will be the same. Endless oceans of slop. How does it work? Nobody knows, but it somehow just kind of does. Kek
>>110280335>>how do you get out of this situation?>open up email clientto: teamleader@work, cc: everyone at worksubject: funny cat memes lolhttps://youtu.be/kqlGTYi_Qco?t=17
to: teamleader@work, cc: everyone at worksubject: funny cat memes lolhttps://youtu.be/kqlGTYi_Qco?t=17
>>110280342>>110280335how long until you niggas start burning incense while praying to the machine spirit and go on quests hunting STCs
>>110280335Ooooh someone didn't set Claude to Max effort and spend all their tokens on linting duplicate logic, dead code, conflicting functions (etc) every week or so and have it maintain a DEVELOPER.md file outlining all the logic flows for the humans. Now they pay the price.
>>110280669we already do
>>110280335>keep re-implementing helpersEspecially Claude is guilty of this bullshit.AI textwalls are tiresome and so is AI code.
>>110280335yeah there's gonna be a shock period where people have to wade through mental molasses to understand their codebase again. this is what a good IDE and a documentation generator are for.
>>110280335I follow a slop project, that advertises its discord prominently, and every single day i compile github bug report statistics and post it without comment.It makes them extremely angry.Bugs get reported 5 times faster than fixed. You watch them desperately tard wrangle their slopbots, just to go in circles and introduce more bugs faster.They are slopping at high speed and the situation keeps getting worse and worse.
>>110280756Must not be a Go or Python project.
>>110280335skill issue>>110280669I was doing that before AI
>>110280335It's happening. Pic related.This is why you write the code yourself and get the AI to critique it. Not the other way around.
>>110280335I look forward to having 100% machine written code on all devices. meatbag handwritten slop is buggy and inefficient as hell
>>110280756dream on ape
>>110280342>No one understands 100% of how the human body worksare you a retard who thinks there is no creator?
>>110280342In my day people became engineers because they wanted to know how shit works.
>>110280335Their fault for terrible prompting and not keeping the database human readable. Bad engineers will be bad claude engineers.
>>110280669We already passed that stage 20 years ago.
>>110280669That would if anything be an improvement. In that universe the people who get to do that are at least somewhat initiated into how shit works. Here in this hellscape, that is considered "gatekeeping" and is bad.
>>11028033599% of all these situations essentially boil down to 'we never documented anything and never defined rules'. AI really brought the issue of lacking documentation and cohesive rules to light by being overly literal.
>>110280335>be clanker>have limited context window>general wall of text that fill context even faster>basically ruin entire projects
>>110281007I didn't understand a word of the ESL babble you just said even though I'm ESL myself.
>>110280992Yeah, all they had to do in the prompt was say>keep the code human-readable at all timesand it would have worked, right?
>>110280335Tell them just because you can read lord of the rings doesn't mean you can write it, and this applies to code review. You will never catch the same amount of bugs reading code than writing it yourself. the end of the dunning kruger curve for ai assisted programming will be the single line autocomplete.
>>110280335sounds stupidif i join a company then i usually have to work on something that has 10 years or even morethe duplication happens in fully human written systems tooit naturally happens when the system is too large to keep in your head or the people working on the system are too new to tell if something already exists or it is so old nobody recalls
>>110281047You can set coding standards yes. If you don't then you're a retard who shouldn't be coding even without AI.
This is why I always read all the AI generated code and instruct it to make changes which usually end up with 2-3x less LoC and code that is way more reusable and concise. Claude generates abstractions for things that don't need abstractions and duplicate all the code that could be generalized without refactoring. Claude can write code like 100 juniors but it can't design architecture as well as even one senior.
>>110281224>if i join a company then i usually have to work on something that has 10 years or even moreNot every company works on legacy. Many software houses work on smaller, greenfield projects that only have like 1-3 years old codebases.
>>110280669I do this every time something goes wrong with the printer
>>110281009>99% of all these situations essentially boil down to 'we never documented anything and never defined rules'.Even with tons of documentation and rules Claude will write code that is future proof. It just lacks human Intuition into making good robust software. You can't encode all of that into rules, you can try to express main principles, but it won't 100% guide your agent to use right abstractions at the right time and design sane software architecture.I spent last 2 weeks trying to do just that. My docs and skills and like 3 times more lines of text that there was lines of code in the project boilerplate. And Claude still manages to find ways to write shitty code. You can't just boil down all of human' ingenuity into a text document.
>>110281329>Claude will write code that is future proof.*Will not
>>110281329>And Claude still manages to find ways to write shitty codeIs it shitty code or code you don't agree with?
>>110280981Okay, but literally nobody here mentioned the engineers in any capacity, so why the fuck are you bringing that up?Oh! LOL, you meant software "engineers", kekThere is a reason why that isn't a real engineering discipline and has never been one, buddy.
>>110281047>Make Javadoc comments before every class briefly explaining purpose, inputs, and outputs. Make inline comments before each functional block of code that explain what’s happening. womm
>>110280335easy make ai go read it
>>110280964Retard
>>110280335just tell claude to make the code as retarded and prone to breaking as possible.
>>110280342>but it works anyway.and gets cancer and dementia, but that's cool, right?
>>110280696This is what rewrites from scratch are for.
>>110281387Shitty for the reasons mentioned here >>110281285
>>110280756Artcraft?
>>110281585DNA is basically nature's version of vibe coded slop, yes.
>>110281405You forgot making a file explaining every single function in the project
>>110281285That's how anthropic want people to use ai btw
>>110280335this whole thing may read like cope, and honestly tt may be soon to tell, maybe letting a swarm of agents loose on a fucked up codebase will soon be enough to unfuck it; at least do a "clean room" re-implementation using the original program as a source of truthbut this is essentially what we're already observing in my company. paired with us now having spend limits (after months of tokenmaxxing)I am still making preparations for being replaced, thoughbeit
>>110280342How many boosters did you get?
>>110281801every day we get closer to becoming adeptus mechanicus
>>110281757yfw we are all just mother nature's timeslops
>>110281224>if i join a company then i usually have to work on something that has 10 years or even more>the duplication happens in fully human written systems tooThe issue that the ai got to that point in under a year.
